Streaming data at the IRIS DMC - PowerPoint Presentation

Slide1

Streaming data at the IRIS DMC

Collecting, consolidating and re-distribution

Rick Benson Slide2

The DMC’s Import of Streaming Real Time Data: Processes

8

0 slarchive processes retrieve data from SeedLink Servers

20 orb2orb processes retrieve data from Antelope systems

30 ew2mseed processes retrieve data from Earthworm

WaveServers

5

RRPServer

processes receive data from Edge RRP feeds

An Earthworm system imports data from other earthworm,

reftek

, and

guralp

systems

A

Nanometrics

ApolloServer

imports data from

N

anometrics

systems Slide3

Real-Time Protocols Involved

Ways to Improve Configuration Correctness and Data Integrity:

Reduce

the use of protocols that require detailed

config Ew2mseedNanometrics ApolloReftek rtpdIncrease the use of protocols that deliver native miniSEEDRingserver/slarchiveNEIC rrpSlide4

Real-Time Stations in this region:

http://

www.iris.edu

/

gmap/_REALTIME?minlat=-10&maxlat=50&minlon=61&maxlon=150Slide5

REAL TIME DATA STATUS

Buffer of Uniform Data (BUD)

There are

177 networks

currently submitting data to the DMC, and 4 are Partially or Fully RestrictedThere are really 4 “BUDS”Main BUD with 118 networks2,233 stations, 16,496 channelsBUD Restricted (AF, EC, YN)65 stations, 235 channelsUSRA with 9 networks (AZ, C, CI, MC, N4, PB, TA, US, UU)619 stations, 17,827 channels- TA alone has 55,655 channels

USRA Restricted (XO Flex Array)

50 stations, 420 channels

2918 STATIONS with

~107,927

CHANNELS………..and steadily increasing.Slide6

Purpose of collecting streaming data

Efficiency and automation of the archiving process

Much easier than infrequent data delivery

Rapid data availability for users

Batch requests, event-windowed and streamingAllows automated, near real-time quality controlRoutine measurements in near real-timeAccess to data not available otherwiseFor some data contributors real-time is the preferred or only optionSlide7

Streaming data flow to the DMC archive

BUD

Buffer of

Uniform Data

ARCHIVEBATS

DCC

DCC

DCC

DCC

…

Data holdings:

Seconds to 8 weeks

behind current time*

12-24 hours after

arrival

Data holdings:

24 hours to

40+ yearsSlide8

Buffer of Uniform Data, the ins and outs

BUD

Buffer of

Uniform Data

EarthwormSeedLink

RRP

Reftek

Guralp

Antelope

CD 1.x

Mini-SEED

Wilber3

www.iris.edu/wilber3/

DMC Web Services

Service.iris.edu

SeedLink Export

www.iris.edu/data/dmc-seedlink.htm

NanometricsSlide9

So that’s how data gets

INTO the DMC.

Now, how do we

EXPORT

data in real-time?Slide10

The DMC’s SeedLink export service

SeedLink server, publically accessibleHost:

rtserve.iris.washington.edu

Port: 18000

All open data in the BUD is available via SeedLink with minimal added latency. Median latency: 40 Hz => 12 seconds, 1 Hz => 160 secondsSlide11

The SeedLink protocol can be summarized as a simple, ASCII-based, data selection phase followed by the streaming of data packets from the server. SeedLink packets are composed of a small header followed by a 512-byte Mini-SEED record (data only SEED). The negotiation phase allows the client to request only specified data from the server for each selected data stream. A data stream is defined by a network and station code pair. 

By utilizing sequence numbers for each packet in a data stream the SeedLink protocol allows for connections to be resumed, eliminating most data gaps. The ability to resume data streams is primarily dependant on how much data, time-wise, the remote SeedLink has in its buffer. 

Special, out-of-band packets created by a

seedlink

server and recognized by libslink are used to communicate server details to clients and to implement keep-alive packet swapping. These special INFO packets are XML formatted data embedded in Mini-SEED comment records. The protocol allows for two different modes of data transmission, uni-station and multi-station modes. Uni-station mode operates by transmitting a single data stream (data from a single station) through one network connection. In this mode the data stream does not need to be specified by the client as it is implied by the internet address and port. Multi-station mode operates by transmitting multiplexed data streams (data from multiple stations) through a single network connection. Almost all connections are negotiated as multi-station, even if only a single station is requested; uni-station mode, for most publicly accessible servers is deprecated.

 SeedLink was originally created as the transport layer for the SeisComP package developed by GEOFON.

Protocol DetailsSlide12

IRIS Supported SeedLink clients:

http://www.iris.edu/data/dmc-seedlink.htm

Slide13

SeedLink export data flow

BUD

512-byte

Mini-SEED

ScannerScanner

Scanner

Scanner

…

Ringserver

~8 hours of data

SeisComP

slarchive

slink2orb

slink2ew

NAQS

…

2,918 stations

33,000+ channelsSlide14

Snapshot of SL client connections: Q1-2014

An average of 600-700 connections active at all times

Over 65 gigabytes per day stream outSlide15

The IRIS Turnkey SeedLink Server:

Called “Ringserver”Slide16

Ringserver: a stand alone SeedLink server

The IRIS DMC’s SeedLink implementation is being released to promote data exchange. Any data center creating 512-byte Mini-SEED records can have a SeedLink server.

Source code

:

https://seiscode.iris.washington.edu/projects/ringserver SeedLink server configuration instructions:https://seiscode.iris.washington.edu/projects/ringserver/wiki/How_to_configure_ringserver_as_a_SeedLink_streaming_serverSlide17

Highlights of

ringserver implementation

Stateful and self-correcting operation

Extremely scalable threaded architecture

Ring buffer is First In First Out (FIFO)Buffer size & number of clients limited only by hardwareDynamic configuration file optionsControl access by IP address, limit access at stream levelSeedLink 3.1 plus Network and Station wildcardingComprehensive transfer loggingRuns on Linux, Mac OSX and Solaris (32 and 64-bit )Integrated MiniSEED file system scannerSlide18

Ringserver related software

dalitool – General purpose ringserver query

Report server ID, version, general status, transfer rate

List connected clients and associated stats

Data stream inspectiondali2liss – Create LISS servers from ringserver streamsslink2dali – Send SeedLink streams to ringserver (beta)mseedscan2dali – Alternate Mini-SEED scannerSlide19

Ringserver conclusions

Scales to huge numbers of stations and channels

Has run uninterrupted at the IRIS DMC for many months

Dynamically reconfigurable to limit downtime

Tracks and logs comprehensive user statisticsringserver created after attempts at using other systemsSlide20

Submitting Non-Real Time Data to IRIS

miniseed2dmc

: A dedicated program for submitting batches of Mini-SEED data directly to the IRIS DMC.

Robustly transmits data to the DMC, tolerant of network disconnects and transmission restarts

Designed to transmit very large data setsReports transmission summaryDataless SEED metadata must be supplied separatelyDistributed as source code, a C compiler is required

Coordination with the IRIS DMC is required before submitting data.Slide21

Sources of Data: Specific SolutionsSlide22

IRIS DMC Earthworm: Data to Archive

As part of our Earthworm system, the IRIS DMC

runs an

ew2ringserver=>

ringserver=>SeedLink server and an slarchive process internally to write data imported by our earthworm into our standard archival of miniSEED.We suggest that, where possible, network operators running earthworm also run these processes to best serve data to the DMC.Slide23

Network Operators Can Run An

Integrated Earthworm to SeedLink Server using ‘Ringserver’

Serves buffered Earthworm data to the DMC and and any other data clients via the widely used standard SeedLink protocol (

tcp

, miniSEED)Can be controlled with earthworm startstop and write log files to earthworm log file areaConvenient way to get miniSEED data out of earthworm for local use tooSlide24

It works like this:

ew2ringserver->ringserver->SL clients

ew2ringserver is available in earthworm v7.7 and is a standard earthworm module. Ew2ringserver exports data to a

ringserver

.ringserver is a stand alone C program that runs on unix/linux* and serves data via the seedlink protocol. Standard SL (SeedLink) clients can access ringserver data. ringserver has features that allow it to easily integrate with earthworm

Detailed instructions to run ew2ringserver->

ringserver

are available at:

https

://seiscode.iris.washington.edu/projects/ew2ringserver/

wiki/A_SeedLink_server_for_Earthworm

*If

you are running earthworm on Windows but have access to a

unix

system, you can still run ew2ringserver to export data to a

ringserver

running on the

unix

system. However, the

ringserver will be running outside earthworm startstop control so is less well integrated.

ISlide25

To Run a Ringserver…

D

ownload, compile, and place the resulting executable

ringserver

in the earthworm user’s PATH. Get ringserver here:https://seiscode.iris.washington.edu/projects/ringserverPrior to running ringserver, you will need to create a directory that will be used by ringserver as the buffer area for dataIf needed open a hole in your firewall to allow data client to open the seedlink port (defined in the ringserver.d file, typically 18000)Slide26

Ew2ringserver

Add an ew2ringserver instance to your earthworm system in the standard earthworm module way.

In the ew2ringserver.d file, the

ringserver’s

IP address and data input port (typically port 16000) is specified by this parameter: RSAddress localhost:16000Slide27

To Integrate Ringserver with Earthworm

You Need to:

Put a

ringserver.d

file in your earthworm params dir like:---RingDirectory /data/my_ringserver_dataDataLinkPort 16000SeedLinkPort 18000ServerID “MySeedlinkServer”---

Put an entry in the

startstop.d

file like:

…

Process

"

ringserver

/

my_ew_run_dir/

params/ringserver.d

-STDERR"

Class/Priority TS 0

…Slide28

Useful

Auxilary Programs to Monitor a Ringserver Seedlink Server

s

linktool

– use to to confirm that a seedlink server is up, that you can connect to it, and that it is serving data.dalitool – use to view clients connected to a ringserver and moreslarchive – use to download data from a SeedLink serverSlide29

Use

dalitool to monitor a ringserver

Specific to

ringserver

, not for generic seedlink server Dalitool can be used to inspect currently connected clients of a ringserverSource code for dalitool can be downloaded from:http://www.iris.edu/pub/programs/ringserver/Slide30

View clients of a

ringserver using dalitool:

dalitool

–ff –C

your.ringserver.ip.address:16000…ops1.iris.washington.edu [192.168.166.67:56328] [SeedLink] SeedLink Client 2014-05-21 20:58:22.983494 Packet 1598812 (2014-05-21 20:45:39.069500) Lag 0%, 1.8 seconds TX 185 packets 0.0 packets/sec 94720 bytes 0.0 bytes/sec RX 0 packets 0.0 packets/sec 0 bytes 0.0 bytes/sec Stream count: 46

Match: ^IU_.*_.*$

Reject:

…

The above example connection displayed by

dalitool

is from running the command ‘

slinktool

-p -S "IU_*" rtserve:

18000’Slide31

Use

slinktool to query any seedlink server

slinktool

–h for help/usage

slinktool –Q localhost:18000 – for a list of streams and time spans of data currently available on the SeedLink server running on the local machine, port 18000Slinktool –L localhost:18000 – for a list of stations served by the SeedLink server running on the local machine, port 18000Slide32

slarchive

Slarchive can be used to retrieve data from a SeedLink server, write it to disk specify “stream selection”, etc

-S streams Define a stream list for multi-station mode

'stream' is in NET_STA format, for example:

-S "IU_KONO:BHE BHN,GE_WLF,MN_AQU:HH?.D"Slarchive –h for help/usageDownload slarchive from here:http://www.iris.edu/dms/nodes/dmc/software/downloads/slarchive/Slide33

The DMC’s

Export of Streaming Real Time Data

We Use Ringserver as our SeedLink Server

Typically there are approx 650+ connections to our

SeedLink server at all timesReal time data is also available from Web services and Breq_fast Real time data is used for some of the IRIS DMC’s data productsSlide34

The DMC runs a publicly accessible SeedLink server on the following

host and port:host:

rtserve.iris.washington.edu
port

: 18000 (default SeedLink port)

All open data that the DMC receives in real-time is available via this SeedLink server.Data arriving at the DMC more than 48 hours behind real-time are not exported via SeedLink.Usage Restrictions:
Users are welcome to any data available via the server as long as client actions do not inhibit our capability to deliver data to other users.To view all currently available real-time stations:http://www.iris.edu/mda/_REALTIMEhttp://www.iris.edu/gmap/_REALTIME Currently 2918 stations (On July 3, 2014) and >32,647

chans

Real-Time Data from IRIS DMCSlide35

To improve communication about real time data issues, the DMC manages a real time mailing list.  When problems with real time data flow are observed, messages will be posted to this list as the problems are discovered.  Solutions to the problems or information regarding resumption of real time data feeds will also be posted to this list.

You can subscribe to this list at 


http://www.iris.washington.edu/mailman/listinfo/iris-rtfeeds

Real-Time mailing list